RT-AC5300 design problem in power consumption?

[manilla]

I'm really happy with my Asus RT-AC5300 router in functionality, but

can someone understand, that this router beeing configured as "ACCESS POINT" and WiFi switched off, still consumes 150mA, just for keep on working the Gigabit Switch?

manilla

 

[L&LD]

I'm really happy with my Asus RT-AC5300 router in functionality, but

can someone understand, that this router beeing configured as "ACCESS POINT" and WiFi switched off, still consumes 150mA, just for keep on working the Gigabit Switch?

manilla

150mA at what voltage (19V DC output) is less than 3W. You can't expect it to work otherwise?

Some power is still required to operate the switch circuit(s).

 

[RMerlin]

I'm really happy with my Asus RT-AC5300 router in functionality, but

can someone understand, that this router beeing configured as "ACCESS POINT" and WiFi switched off, still consumes 150mA, just for keep on working the Gigabit Switch?

manilla

There isn't just as witch, the router also has a dual core CPU running an operating system. A router has far more components than an unmanaged switch.

 

[i_max]

I have a 8 port gigabit switch, I can trade for that router if you'd like

 

[manilla]

150mA at what voltage (19V DC output) is less than 3W. You can't expect it to work otherwise?

Some power is still required to operate the switch circuit(s).

sorry for not having mentioned the voltage,

no,
I wouldn't be surprised for a value of 150mA at (19V Output), configured as Access Point and WiFi switched off,
the RT-AC5300 actually consumes in this status 150 mA at (220V AC Output), that is like a 33-Watt light bulb ;-)
...and configured as Access Point and WiFi switched on ---> 200mA = 44 Watt
...and configured as Router and WiFi switched on ---> ??? (didn't measure yet)

@RMerlin,
yeah, I know AC5300 isn't just a Switch, but todays advanced CPU designs would actually allow to handle different Status much better in power consumption!
Think about our modern Notebooks these days beeing able to work for 10 to 15 hours with just one battery Charge.

manilla

 

[L&LD]

sorry for not having mentioned the voltage,

no,
I wouldn't be surprised for a value of 150mA at (19V Output), configured as Access Point and WiFi switched off,
the RT-AC5300 actually consumes in this status 150 mA at (220V AC Output), that is like a 33-Watt light bulb ;-)
...and configured as Access Point and WiFi switched on ---> 200mA = 44 Watt
...and configured as Router and WiFi switched on ---> ??? (didn't measure yet)

@RMerlin,
yeah, I know AC5300 isn't just a Switch, but todays advanced CPU designs would actually allow to handle different Status much better in power consumption!
Think about our modern Notebooks these days beeing able to work for 10 to 15 hours with just one battery Charge.

manilla

There are inefficiencies from converting from AC power to DC. If you are measuring at the wall, you are also measuring how (inefficient) the adaptor is at it's lowest and most inefficient power draw. I highly doubt the RT-AC5300 is using anywhere close to 33W in that usage mode (it may be half or less), even if it is rated for 65W maximum (well, the power adaptor is; 19V @ 3.42A).

Routers and laptops processors are not equivalent. Laptop (and desktop) processors have many features to extend battery life when they're not being actively used (but still 'on'). Speedstep and enhanced speedstep is used along with many other technologies to achieve that goal of very long battery life.

A router though does not use anything like that (today). It is running at full frequency at all times. No power savings enabled at all.

If you want to accurately see what the power consumption of the RT-AC5300 is, you need to monitor the voltage and amperage of the DC input lines on the router. That will take out the inefficiencies of the AC adaptor/converter and give you an actual wattage that the router needs in the configuration you have it in. Of course, you do pay for the AC adaptor and router combination in electricity, but again, at such low loads, the inefficiencies of the power supply 'should' be ignored, imo.

Just one opinion, of course.

 

[TonyH]

sorry for not having mentioned the voltage,

no,
I wouldn't be surprised for a value of 150mA at (19V Output), configured as Access Point and WiFi switched off,
the RT-AC5300 actually consumes in this status 150 mA at (220V AC Output), that is like a 33-Watt light bulb ;-)
...and configured as Access Point and WiFi switched on ---> 200mA = 44 Watt
...and configured as Router and WiFi switched on ---> ??? (didn't measure yet)

@RMerlin,
yeah, I know AC5300 isn't just a Switch, but todays advanced CPU designs would actually allow to handle different Status much better in power consumption!
Think about our modern Notebooks these days beeing able to work for 10 to 15 hours with just one battery Charge.

manilla

Used Amprobe to measure the AC input side consumption? Isn't adapter getting warm?, lost energy as heat. Switching power supply is lot less than 100% efficient. Real consumption should be measured
at 19V DC output side. There are two cpus in the unit. Even tho WiFi radios are disabled rest of unit
is still drawing juice. How much capacity(in Amp/hours) laptop batteries have? Laptop chargers are
rated something like 45 to 135Watts at outputput side. My little 14" Alienware laptop is rated at 2.5Amp. on universal input voltage. That is ~250W. It depends what laptop you are referring too.

 

[RMerlin]

yeah, I know AC5300 isn't just a Switch, but todays advanced CPU designs would actually allow to handle different Status much better in power consumption!
Think about our modern Notebooks these days beeing able to work for 10 to 15 hours with just one battery Charge.

Cortex A9 ain't a very recent design, and Broadcom didn't implement any clock throttling in their design (I assume it's for latency reasons).

150 mA, taking into account what might be around 75% power efficiency (Asus's recent power adapters are much better than the old ones, but I doubt they're in the 80-85% ratio for a 220V user) doesn't seem much to me for a dual core 1.4 GHz CPU. It's also a 28 nm litography if I remember.

 

[TonyH]

There are inefficiencies from converting from AC power to DC. If you are measuring at the wall, you are also measuring how (inefficient) the adaptor is at it's lowest and most inefficient power draw. I highly doubt the RT-AC5300 is using anywhere close to 33W in that usage mode (it may be half or less), even if it is rated for 65W maximum (well, the power adaptor is; 19V @ 3.42A).

Routers and laptops processors are not equivalent. Laptop (and desktop) processors have many features to extend battery life when they're not being actively used (but still 'on'). Speedstep and enhanced speedstep is used along with many other technologies to achieve that goal of very long battery life.

A router though does not use anything like that (today). It is running at full frequency at all times. No power savings enabled at all.

If you want to accurately see what the power consumption of the RT-AC5300 is, you need to monitor the voltage and amperage of the DC input lines on the router. That will take out the inefficiencies of the AC adaptor/converter and give you an actual wattage that the router needs in the configuration you have it in. Of course, you do pay for the AC adaptor and router combination in electricity, but again, at such low loads, the inefficiencies of the power supply 'should' be ignored, imo.

Just one opinion, of course.

If knit picking within the router 19V DC has to be down to 5V DC again using regulators incurring power loss, probably there is need fo 12V DC too, and so on and on.

 

[manilla]

>>Cortex A9 ain't a very recent design, and Broadcom didn't implement any clock throttling
>>in their design (I assume it's for latency reasons).

that's the point!

 

[L&LD]

If knit picking within the router 19V DC has to be down to 5V DC again using regulators incurring power loss, probably there is need fo 12V DC too, and so on and on.

Yes, you are nitpicking.

That internal conversion (and associated losses) is still part of the router's power envelope.

 

[RMerlin]

>>Cortex A9 ain't a very recent design, and Broadcom didn't implement any clock throttling
>>in their design (I assume it's for latency reasons).

that's the point!

Your thread title said "design problem". It's not a design problem, it's what is expected out of that specific hardware used by this router. You will find the same thing with any other router using similar hardware.

 

[dlandiss]

150 mA, taking into account what might be around 75% power efficiency (Asus's recent power adapters are much better than the old ones, but I doubt they're in the 80-85% ratio for a 220V user) doesn't seem much to me for a dual core 1.4 GHz CPU.

It is also true that clamp-on current meters do not account for power factor--either from phase angle or non-sinusoidal waveforms. The true power can be quite a bit less than volts times amps (scalar).

 

[sfx2000]

Cortex A9 ain't a very recent design, and Broadcom didn't implement any clock throttling in their design (I assume it's for latency reasons).

At last count, there's at least four different Cortex-A9 releases across 40nm and 28nm - so it could be any one of them - surprisingly enough, Broadcom is still rolling out 40nm chips - the 2709 is Cortex-A53 based, and it's still on 40nm, which is probably due to the on-board VideoCore engine (gpu).

(It's the VideoCore, by the way, why we might never see 64-bit linux on the Pi3's - it's a 32bit part)

Pretty much impossible to tell unless someone from an outfit like Chipworks decaps the chip and takes a look...

 

[RMerlin]

At last count, there's at least four different Cortex-A9 releases across 40nm and 28nm - so it could be any one of them

The information I found was specific to the BCM4709C0, which was labeled as being a 28 nm part.

 

[sfx2000]

The information I found was specific to the BCM4709C0, which was labeled as being a 28 nm part.

Nice... and one might be able to discover which ARMv7 (Cortex-A9) minor release is in /proc/cpuinfo perhaps...

 

[RMerlin]

Nice... and one might be able to discover which ARMv7 (Cortex-A9) minor release is in /proc/cpuinfo perhaps...

The information filled by Broadcom's kernel code is half garbage. /proc/cpuinfo still identifies even the latest BCM4709C0 as being a "Northstar prototyype"... Northstar was the codename of their first generation ARM HND (i.e. what the RT-AC56U was based on).

The rest is mildly useful...

Features    : swp half thumb fastmult edsp
CPU implementer    : 0x41
CPU architecture: 7
CPU variant    : 0x3
CPU part    : 0xc09
CPU revision    : 0

 

[sfx2000]

Hard to tell - and this is an older kernel that Broadcom supplies that might not show enough info on the HW at that level - with JTAG one might be able to get deep into the registers and peek there...

Some background info from ARM on the multiple A9 variants (I should have added this earlier)

Product revisions

 

[sfx2000]

Should also point out that while the BCM4708/4709, as RMerlin suggests above, do not support dynamic clocks (not needed in this application), they have enabled power gating on the A9's and the ports on the switch functional element - which puts the SoC at around 100ma when idle - which is pretty good - that additional 50ma draw is probably the PHY complex over to the ethernet ports - as one has to keep them lit up, even with green ethernet (or energy efficient ethernet)...

 

[sfx2000]

If knit picking within the router 19V DC has to be down to 5V DC again using regulators incurring power loss, probably there is need fo 12V DC too, and so on and on.

Actually this is a common approach - bringing in VDC at a higher voltage - and then using a switching power complex on the circuitboard to break out the various DC rails - 5VDC, 3.3VDC, and commonly 1.8VDC.

Details - input DC comes in, and gets converted to very high frequency AC, and then get split out across the different voltage levels and converted back to DC at those level...

One brings in the DC at 19V, and that's basically for efficiency purposes, and less current - power in (and recall that power is voltage*current) and power out of the complex is pretty close to the same, on the order of 80-90 percent - there's a bit of switching loss and some heat (very little)...

So considering that the SoC is pulling about 100ma in idle state, and one considers all the peripherals, 150ma at the wall is pretty darn efficient if you ask me...